Korean relative clause constructions: Conspiracy and pragmatics

28
This article was downloaded by: [University of Arizona] On: 09 August 2012, At: 02:21 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Australian Journal of Linguistics Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/cajl20 Korean relative clause constructions: Conspiracy and pragmatics Jae Jung Song a a Linguistics English Department, University of Otago, P.O.Box 56, Dunedin, New Zealand Version of record first published: 14 Aug 2008 To cite this article: Jae Jung Song (1991): Korean relative clause constructions: Conspiracy and pragmatics, Australian Journal of Linguistics, 11:2, 195-220 To link to this article: http://dx.doi.org/10.1080/07268609108599462 PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/ terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising

Transcript of Korean relative clause constructions: Conspiracy and pragmatics

This article was downloaded by: [University of Arizona]On: 09 August 2012, At: 02:21Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number:1072954 Registered office: Mortimer House, 37-41 Mortimer Street,London W1T 3JH, UK

Australian Journal ofLinguisticsPublication details, including instructions forauthors and subscription information:http://www.tandfonline.com/loi/cajl20

Korean relative clauseconstructions: Conspiracyand pragmaticsJae Jung Song aa Linguistics English Department, University ofOtago, P.O.Box 56, Dunedin, New Zealand

Version of record first published: 14 Aug 2008

To cite this article: Jae Jung Song (1991): Korean relative clause constructions:Conspiracy and pragmatics, Australian Journal of Linguistics, 11:2, 195-220

To link to this article: http://dx.doi.org/10.1080/07268609108599462

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private studypurposes. Any substantial or systematic reproduction, redistribution,reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or makeany representation that the contents will be complete or accurateor up to date. The accuracy of any instructions, formulae, and drugdoses should be independently verified with primary sources. Thepublisher shall not be liable for any loss, actions, claims, proceedings,demand, or costs or damages whatsoever or howsoever caused arising

directly or indirectly in connection with or arising out of the use of thismaterial.

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

© Australian Journal of Linguistics 11 (1991), 195-220. Printed in Australia.

KOREAN RELATIVE CLAUSE CONSTRUCTIONS:CONSPIRACY AND PRAGMATICS

Jae Jung Song*

1. INTRODUCTION

Keenan and Comrie (1977) put forward the Accessibility Hierarchy(hereafter AH) in the form of:1-2

(1) AHSU > DO > 10 > OBL > GEN > OCOMP(where > means 'is more accessible than')

The AH reflects the relative accessibility to relativization of NP positions.3

The thrust of their theory is that if a language can form relative clauses on agiven position, it can also form relative clauses on all positions higher on theAH. They (1977:68) also claim: (a) a language must have a primaryR(elative) C(lause) forming strategy, which is used to relativize subject NPs;and (b) it need have no other (non-primary) strategies.

In particular, they (1977:74) claim on the basis of Tagashira (1972) that inKorean the primary RC forming strategy - what Comrie (1981:144) calls'the gap strategy' in this case - applies to all NPs on the AH down to andinclusive of oblique NPs (e.g. (2); relative clauses are enclosed in squarebrackets):4

* I am indebted to Keith Allan and Barry Blake for reading earlier drafts of this paper and offeringinsightful comments. Special mention must be made of Graham Mallinson, who initiallyencouraged me to work on Korean relative clauses and also provided useful suggestions on the firstdraft. I have benefited from the useful comments provided by the two anonymous referees of AJL.Needless to say, none of these scholars are responsible for the errors that I may have failed toeliminate and the way I have made use of their valuable contributions.

1 To avoid potential confusion, in what follows, I will capitalize the first letter of each of the wordsof the term, accessibility hierarchy, when I refer to Keenan and Comrie's accessibility hierarchy(i.e. Accessibility Hierarchy). On the other hand, I will employ no capitalization when I refer toNa's (1986) accessibility hierarchy (i.e. accessibility hierarchy).

2 The following abbreviations are used in the paperABL = Ablative, AH = Accessibility Hierarchy, ALL = Allative, COM = Comitative, CONJ =Conjunction, DO = Direct Object, GEN = Genitive, HON = Honorific, IMP = Imperative, IND =Indicative, INST = Instrumental, IO = Indirect Object, LOC = Locative, NOM = Nominative, OBL= Oblique, OCOMP = Object of Comparison, PF = Phonological Filler, PRES = Present, PST =Past, Q = Question, RC = Relative Clause, REL = Relative, SU = Subject, TOP = Topic.

3 Comrie (1975, 1976) attempts to extend the hierarchy to the syntax of causative constructions.However, see Song (1991) for arguments against the extension and an alternative interpretation ofthe phenomenon.

4 The element glossed as REL -n ( or -(n)un) is not really a relative pronoun, but it is some relativeelement that also carries tense. See Tagashira (1972) for more detail. I have changed some propernames in others' Korean data.

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

JAE JUNG SONG

(2) [kiho-ka ku kay-lul ttayli-n] maktaykiKeeho-SU the dog-DO beat-REL stick'the stick with which Keeho beat the dog'

They further note that Korean has a non-primary strategy of pronounretention that is put to use when the genitive NP is relativized (e.g. (3)):

(3) [caki-uy kay-ka chongmyengha-n] ku salamhe-of dog-SU smart-REL the man'the man whose dog is smart'

According to Keenan and Comrie, then, the cutoff point for the gap strategyin Korean is the OBL. However, Tagashira (1972) clearly shows thatKorean relative clause syntax is not so simple and straightforward as Keenanand Comrie (1977) depict it in their widely known cross-linguistic work.For instance, Tagashira (1972:217-219) notes that some oblique NPs cannotbe relativized by way of the primary RC forming strategy or the non-primaryRC forming strategy.

Recently, Matsumoto (1990), following Na (1986), concludes that therole of pragmatics in the generation and construal of relative clauses inKorean is not so significant as in Japanese; in effect, Matsumoto claims thatthe accessibility hierarchy of syntactic, semantic, and pragmatic rolesproposed by Na is valid for Korean relative clauses.

In this paper I will discuss the 'conspiracy' (Croft (1990)) in Koreanrelative clauses: certain 'devices' already available in the grammar -promotion, conjunction, and the use of certain adverbs - are called into therelativization process so that otherwise unrelativizable oblique NPs can berelativized using the primary strategy. To the best of my knowledge, thesedevices save for the promotion device have not been well documented in theliterature on relative clause formation. Further, I will argue that Na'saccessibility hierarchy is too strong for Korean relative clauses: the role ofpragmatics in Korean relative clauses is as significant as in Japanese relativeclauses. Consequently, I hope to clear up the misunderstandings ormisrepresentations of Keenan and Comrie (1977) and Matsumoto (1990)concerning the Korean relative clause.

2. KOREAN RELATIVE CLAUSE: RC FORMING STRATEGIES

Korean is a typical SOV language.5 For instance, postpositions are used tomark core grammatical relations and oblique relations; the auxiliary alwaysfollows the main verb. As already illustrated in (2) and (3), in Korean therelative clause precedes the head NP. Korean uses the gap type ofrelativization as its primary RC forming strategy: the NP relativized on is

5 Greenberg (1966:109) lists Korean as a language of Type 23, the characteristic features of whichare: SOV, Postpositions, Genitives and Adjectives preceding Head Nouns.

196

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

KOREAN RELATIVE CLAUSE CONSTRUCTIONS

'deleted' along with its case marker or postposition. The relative clause in(4) can be related to the sentences in (5) ((5.a) is 'embedded' in (5.b)):

(4) kiho-ka [koyangi-lul mu-n] kay-lul ttayli-ess-taKeeho-SU cat-DO bite-REL dog-DO hit-PST-IND'Keeho hit the dog that bit the cat.'

(5) a. kay-ka koyangi-lul mul-ess-tadog-SU cat-DO bite-PST-IND"The dog bit the cat.'

b. kiho-ka kay-lul ttayli-ess-taKeeho-SU dog-DO hit-PST-IND'Keeho hit the dog.'

The subject NP in (5.a) is relativized in (4): the NP itself and the subjectpostposition -ka are eliminated in the corresponding relative clause in (4).The gap strategy applies further down to oblique NPs such as locative,allative, ablative, instrumental, and comitative. The following examplesillustrate these possibilities. Note that the (b) sentences represent the non-relative clause equivalents to the relative clauses in (a).

(6) LOCATIVEa. [yenghi-ka chum-ul chu-n] kongwen

Yonghee-SU dance-DO dance-REL park'The park in which Yonghee danced'

b. yenghi-ka kongwen-eyse chum-ul chu-ess-taYonghee-SU park-LOC dance-DO dance-PST-IND'Yonghee danced in the park.'

(7) ALLATTVEa. [kiho-ka o-n] hocu

Keeho-SU come-REL Australia'Australia, to which Keeho came'

b. kiho-ka hocu-eyllo o-ass-taKeeho-SU Australia-ALL come-PST-IND'Keeho came to Australia.'

(8) ABLATIVEa. [kiho-ka ttena-n] hocu

Keeho-SU leave-REL Australia' Australia, from which Keeho came '

b. kiho-ka hocu-eyse ttena-ass-taKeeho-SU Australia-ABL leave-PST-IND'Keeho came from Australia.'

197

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

JAE JUNG SONG

(9) INSTRUMENTALa. [kiho-ka ku kay-lul ttayli-n] maktayki

Keeho-SU the dog-DO hit-REL stick'the stick with which Keeho hit the dog'

b. kiho-ka maktayki-lo ku kay-lul ttayli-ess-taKeeho-SU stick-INST the dog-DO hit-PST-IND'Keeho hit the dog with the stick.'

(10) COMTTATIVEa. [kiho-ka kyelhonha-n] ku yein

Keeho-SU marry-REL the lady'the lady whom Keeho married'

b. kiho-ka ku yein-kwa kyelhonha-ess-taKeeho-SU the lady-COM marry-PST-IND'Keeho married the lady.'

However, when it comes to genitives, a different (i.e. non-primary) RCforming strategy is called for: the relative clause must contain the pronominalcounterpart of the head NP, i.e. resumptive pronoun.

(11) GENITIVEa. [caki-uy sensayng-nim-i chongkak-i-si-n] yenghi

self-GEN teacher-HON-SU bachelor-is-HON-REL Yonghee' Yonghee, whose teacher is a bachelor'

b. yenghi-uy sensayng-nim-i chongkak-i-si-taYonghee-GEN teacher-HON-SU bachelor-is-HON-IND' Yonghee's teacher is a bachelor.'

And this non-primary strategy is restricted to genitive NPs. Keenan andComrie (1977:74), thus, claim that in Korean the cutoff point of the gapstrategy is the oblique NP.

The Korean relative clause is characterized as follows in Keenan andComrie (1977:78):

(12) Korean SU DO 10 OBL GEN OCOMP1. prénom. - case + + + +2. prénom. + case +

N.B. 1 represents the primary strategy (i.e. gapping); 2 represents the non-primary strategy (i.e. pronoun retention); + indicates that the strategy appliesgenerally to that NP position; - indicates that it does not; prénom, indicatesthat the relative clause precedes the head NP; + case represents a case codingstrategy, whereas - case represents a non-case coding strategy.

Keenan and Comrie (1977:74) appeal to Tagashira (1972) in support oftheir conclusion in (12). However, although it may be true that all locativeNPs can be relativized using the gap strategy, Tagashira (1972:217-219)

198

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

KOREAN RELATIVE CLAUSE CONSTRUCTIONS

clearly demonstrates that not all oblique NPs can be relativized in this way.Consider the following examples (note that the ungrammatical relativeclauses in (a) are related to the sentences in (b) via the gap strategy):

(13) ABLATIVESa. *[kiho-ka o-n] hocu

Keeho-SU come-REL Australia'Australia, from which Keeho came'

b. kiho-ka hocu-eyse o-ass-taKeeho-SU Australia-ABL come-PST-IND'Keeho came from Australia.'

(14) ALLATIVE?a. *[kiho-ka ttena-n] hocu

Keeho-SU leave-REL Australia'Australia, for which Keeho left'

b. kiho-ka hocu-lo ttena-ass-taKeeho-SU Australia-ALL leave-PST-IND'Keeho left for Australia.'

(15) INSTRUMENTALa.*[kiho-ka ka-n ku cha]

Keeho-SU go-REL the car'the car in which Keeho went'

b. kiho-ka ku cha-lo ka-ass-taKeeho-SU the car-INST go-PST-IND'Keeho went in the car.'

(16) COMITATrVEa. *[kiho-ka san-ey ka-n] sunay

Keeho-SU mountain-ALL go-REL Sunae'Sunae, with whom Keeho went to the mountain'

b. kiho-ka sunay-wa san-ey ka-ass-taKeeho-SU Sunae-COM mountain-ALL go-PST-IND'Keeho went to the mountain.with Sunae.'

Since it is used in order to relativize only genitive NPs, the non-primarystrategy of pronoun retention cannot be used at all to render the aboveungrammatical relative clauses grammatical. I will defer till section 4 furtherdiscussion on why some oblique NPs (e.g. (6)-(10)) can be relativized viathe gap strategy, and why other oblique NPs cannot (e.g. (13)-(16)).

6 This relative clause is grammatical, only if it means 'Australia, to which Keeho came' (cf. ex. (7)above).

7 This relative clause is grammatical, only if it means 'Australia, from which Keeho came' (cf. ex.(8) above).

199

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

JAE JUNG SONG

Further, contrary to Keenan and Comrie (1977:74), not all genitives canbe relativized via the non-primary strategy of pronoun retention; instead theusual gap strategy is used even for some genitive NPs (cf. Tagashira(1972:219-221));8

(17) [Ol*caki-uy caynung-i ttwiyena-n] kiho0/self-GEN talent-SU outstanding-REL Keeho'Keeho, whose talent is outstanding'

In Korean, then, the primary strategy of gapping applies to subject, directobject, indirect object, locative NPs. And this strategy applies down to someother oblique NPs and some genitive NPs. The non-primary strategy ofpronoun retention is restricted to genitive NPs in its application; the genitiveposition is accessible to relativization by way of the two strategies. (12) thenmust be changed to:

(18) Korean SU DO 10 OBL GEN OCOMP1. prénom.-case + + + +/- +/-2. prénom. + case - - - - + / -

3. CONSPIRACY IN KOREAN RELATIVIZATION

What is not certain in the general discussion of relativization in Korean is: isthere any way to relativize those oblique NPs that cannot be relativized viathe primary strategy? As noted above, Tagashira (1972) is well aware of thefact that some oblique NPs cannot be relativized in the usual way, i.e. via thegap strategy. However, it is not stated at all in her work whether or notKorean indeed has any way(s) to relativize these seemingly unrelativizableoblique NPs. In this section I will argue that some oblique NPs that seem tobe directly relativized via the gap strategy in fact have been first promoted toDOs in order to undergo the RC forming strategy. And then I willdemonstrate that apparently unrelativizable oblique NPs can indeed berelativized via the gap strategy, but only in conjunction with some otherdevices.

8 This is not the place for further discussion of the distribution of the two RC forming strategieswith respect to genitive NPs. I just wish to point out that Tagashira's explanation for thedistribution is incorrect Tagashira (1972:219-221) argues that the distinction between 'propertypossessive' and 'instance possesive' underlies the distribution of the two strategies. According toher, property possessive NPs undergo the pronoun retention strategy, whereas instance possessiveNPs undergo the gap strategy. Aside from the difficulty in defining the distinction, thisexplanation is not without problems. For instance, there are property possessive NPs that undergothe gap strategy, e.g.:(i) [sin-i ta talh-un] kiho

shoe-SU all wom=out-REL Keeho'Keeho, whose shoes are all worn out'

Obviously, further research is needed.

200

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

KOREAN RELATIVE CLAUSE CONSTRUCTIONS

3.1 Devices brought into the conspiracy

Keenan and Comrie (1977:68-69,88-89) note that Toba Batak can directlyrelativize subject NPs using one strategy, and IOs, OBLs, and GENs usinganother, and that DOs cannot be directly relativized using any strategy.However, DOs can be relativized by first being promoted to SU NPs; inother words, DOs can be indirectly relativized. Thus all NPs including DOscan be relativized. This is what Croft (1990:197) refers to as 'the conspiracybetween accessibility in relative clause formation and promotion of DOs toSUs'. And this conspiracy satisfies the functional communicativerequirement that it be possible to relativize any NP argument of a relativeclause (Croft 1990:198-199). This prompts one to look for such devices thatmay be exploited in Korean so that the strategy in question may be able torelativize those oblique NPs that it cannot relativize otherwise (e.g. (13),(14), (15), and (16)). Indeed, Korean involves certain interesting devices inthe conspiracy: (a) promotion; (b) conjunction; and (c) adverbs. And I willlook at each of them in turn.

3.1.1 Promotion device

The gap strategy, as shown above, eliminates the coreferential NP in therelative clause along with its postposition. So if the strategy applies to theallative NP in (7.b) and the ablative NP in (13.b)*with hocu 'Australia' as thehead NP (the verb and the subject NP remain the same in (7.b) and (13.b)),their corresponding relative clauses would both end up as the same (cf. (7.a)and (13.a)). But (7.b), not (13.b), can undergo the gap strategy, as thegrammaticality of (7.a) and the ungrammaticality of (13.a) clearly indicate:

(7) ALLATIVEa. [kiho-ka o-n] hocu

Keeho-SU came-REL Australia'Australia, to which Keeho came'

b. kiho-ka hocu-eyllo o-ass-taKeeho-SU Australia-ALL come-PST-IND'Keeho came to Australia.'

(13) ABLATIVEa. *[kiho-ka o-n] hocu

Keeho-SU come-REL Australia'Australia, from which Keeho came'

b. kiho-ka hocu-eyse o-ass-taKeeho-SU Australia-ABL come-PST-IND'Keeho came from Australia.'

201

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

JAE JUNG SONG

In other words, the allative NP, not the ablative NP, can be relativized (seesection 3.1.2 on how (13.b) can be relativized in conjunction with a differentdevice).

Now consider (8) and (14) (note again that the verb and the subject NPremain the same, as in the pair of (7.b) and (13.b)):

(8) ABLATIVEa. [Idho-ka ttena-n] hoot

Keeho-SU leave-REL Australia'Australia, from which Keeho came'

b. kiho-ka hocu-eyse ttena-ass-taKeeho-SU Australia-ABL leave-PST-IND'Keeho came from Australia.'

(14) ALLATIVEa. *[kiho-ka ttena-n] hocu

Keeho-SU leave-REL Australia'Australia, for which Keeho left'

b. kiho-ka hocu-lo ttena-ass-taKeeho-SU Australia-ALL leave-PST-IND'Keeho left for Australia.'

The translation for (8.a) shows that the eliminated NP is an ablative one.This is why (8.a) never means 'Australia, for which Keeho left'; notsurprisingly, the allative NP cannot be relativized in (14.b), as indicated bythe ungrammaticality of (14.a) (see the next section on how (14.b) can berelativized exploiting a different device).

In the pair of (7) and (13), it is the allative NP, not the ablative NP, thatcan be relativized via the gap strategy. On the other hand, in the pair of (8)and (14), it is the ablative NP, not the allative NP, that can be so relativized.Why does the gap strategy discriminate in this way?

In the literature on Korean relativization it has been so far taken forgranted that the gap strategy directly applies to allative NPs as in (7.b) orablative NPs as in (8.b). However, I would like to claim that there is apromotion device involved in the relativization process, a kind similar to thatwhich Keenan and Comrie (1977) discuss in connection with Toba Batak.So the allative NP in (7.b) must be first promoted to a DO to be relativizedvia the gap strategy. Similarly, the ablative NP in (8.b) can be relativized,only because it has already become a DO using the promotion device. Andas will become clear, it depends on the verb whether the allative or theablative NP is to be promoted to a DO.

Indeed, in Korean allative or ablative NPs can be promoted to DOs. Noelaborate grammatical machinery (e.g. verbal affixes registering promotion)is used in the promotion; the NP in question can have its oblique postpositionreplaced with the DO postposition -(l)ul. So with respect to the verb o- in(7.b), it is the allative NP, not the ablative NP, that can be promoted in this

202

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

KOREAN RELATIVE CLAUSE CONSTRUCTIONS

way. This is why in (19), in which the promotion has taken place, only the'allative' reading is allowed:

(19) kiho-ka hocu-lul o-ass-taKeeho-SU Australia-DO come-PST-IND'Keeho came to Australia.', but'*Keeho came from Australia.'

In other words, (19) is related to:

(7) ALLATIVEb. kiho-ka hocu-ey/to o-ass-ta

Keeho-SU Australia-ALL come-PST-IND'Keeho came to Australia.'

But it can never be related to:

(13) ABLATIVEb. kiho-ka hocu-eyse o-ass-ta

Keeho-SU Australia-ABL come-PST-IND'Keeho came from Australia.'

And as has already been shown, it is also the allative NP (i.e. (7.b)), not theablative NP (i.e. (13.b)), that can be relativized. Clearly, it is the allativeNP, not the ablative NP, that is involved in the two distinct phenomena, i.e.promotion in (19), and relativization in (7). This parallelism can be easilyaccounted for if it is assumed that the allative NP in question is firstpromoted to a DO to be relativized via the gap strategy, which freelyrelativizes DOs. On the other hand, the ablative NP in (13.b) cannot berelativized, because it fails to be promoted to a DO in the first place.

In contrast, in the case of the verb ttena- in (8), it is the ablative NP, notthe allative NP, that can be promoted to a DO (indicated by the DOpostposition on hocu 'Australia'):

(20) kiho-ka hocu-lul ttena-ass-taKeeho-SU Australia-DO leave-PST-IND'Keeho left Australia', but'*Keeho left for Australia'

In (20) only the 'ablative', not the 'allative', interpretation is possible; (20) isnot related to (14.b), but to (8.b):

(8) ABLATIVEb. kiho-ka hocu-eyse ttena-ass-ta

Keeho-SU Australia-ABL leave-PST-IND'Keeho left Australia.'

203

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

JAE JUNG SONG

(14).. ALLATTVEb. . kiho-ka hocu-lo ttena-ass-ta

Keeho-SU Australia-ALL leave-PST-IND'Keeho left for Australia.'

And as has already been demonstrated, it is also the ablative NP in (8.b), notthe allative NP in (14.b), that can be relativized. Again, one finds the samekind of parallelism between the promotion in (20), and the relativization in(8): the same oblique NP undergoes the two distinct operations. Once theablative NP in (8.b) has been promoted to a DO as in (20), it can easilyfurther undergo the gap strategy, which is shown to freely relativize DOs.On the other hand, the allative NP in (14.b) cannot be first promoted to aDO, thus failing to undergo the gap strategy.

In sum, when this promotion device is recognized in the relativization ofthe above oblique NPs, it is possible to explain the systematic parallelism thathas been observed between the promotion and relativization phenomena: onlythe kind of oblique NP that is promoted to a DO can further undergo the gapstrategy.

The promotion device is used only for allative and ablative NPs.Therefore, a possible semantic role confusion with the other oblique relationsis completely ruled out; for instance, an instrumental NP cannot be promotedto a DO. But what about the confusion between the allative and ablativeroles? When an allative or ablative NP is promoted to a DO, the explicitoblique postposition is replaced by the DO postposition. Thus, the meaningof the oblique postposition is completely lost in the process; the casedistinction between the allative and ablative is neutralized. However, thisdoes not result in any potential semantic confusion, because, as has beenshown, the promotion is allowed to apply to either allative or ablative NPs,depending on the verb involved.9

9 Givón (1990: 576-578) discusses a similar phenomenon in the English promotional (canonical)passive (as in (i)) and the non-promotional (impersonal) passive (as in (ii)).(i) a. She was seen on the beach.

b. She was given a book.c. *The movie was gone to.d. *The pen was written the letter with.e. *The kids were gone there with a lot.f. *Easily was done it.

(ii) a. They saw her on the beach yesterday.b. They gave Mary a book to read.c. They go to the movies a lot here.& You write letters with a pen.e. One goes there with kids a lot.f. You can break it real easy. '

In the former type of passive in (i), the promotion of non-agents to subjecthood is rigidlyconstrained (as indicated by (i.c), (i.d), (i.e), and (i.f)), whereas the latter type in (ii) does notexhibit such a restricted distribution. Givón attempts to explain the restrictions on thepromotional type of passivization in (i) by arguing that they serve as a kind of compensatorystrategy that helps avoid the semantic role confusion that arises from the case neutralization that isin turn caused by the promotion of the non-agent NP to the subject or topic.

204

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

KOREAN RELATIVE CLAUSE CONSTRUCTIONS

3.1.2 Conjunction Device

Now consider (13) once again:

(13) ABLATIVEa. *[kiho-ka o-n] hocu

Keeho-SU come-REL Australia'Australia, from which Keeho came'

b. kiho-fca hocu-eyse o-ass-taKeeho-SU Australia-ABL come-PST-IND' Keeho came from Australia. '

(13.a) is ungrammatical in the sense of (13.b), because with the verb o- onlyallative NPs can be promoted to DOs in order to become eligible for the gapRC forming strategy. However, this does not mean that Korean lacks a wayto relativize the ablative NP in question. There is another device that isrecruited into the relativization conspiracy: the conjunction device.

In Korean there are two major ways of conjoining clauses: (a) the use ofthe conjunctive marker -ase;w and (b) the use of the conjunctive marker -ko(for more detail see Lukoff (1982:105-106,129-132,159,364-365)). Inorder to avoid potential confusion, I will, where necessary, call the first typethe A conjunction device, and the second type the K conjunction device.And I will look at the way that each of these two conjunction devices isrecruited into the relativization conspiracy.

First, the A conjunction device is used in collaboration with the promotiondevice in order to relativize those allative or ablative NPs that cannot bepromoted to DOs. The conjunctive marker -ase is normally used to denote 'aseries of connected or related actions, conditions, and so on' (Lukoff1982:106), e.g.:

(21) ku-nun anc-ase chinku-lul kitali-ess-tahe-TOP sit=down-CONJ friend-DO wait-PST-IND'As for him, he sat down and waited for his friend.' or'As for him, he waited for his friend, while being seated.'

The conjunction sentence in (21) describes a situation where the actor isinvolved in two closely related activities of sitting and waiting: the actor sitsdown in order to wait. So (21) never indicates that the actor is engaged inone activity at one point in time, and then engaged in the other (unrelated)activity at a later point in time. In other words, the conjunction device isnever used to simply list unrelated activities.

Recall from section 3.1.1 that in contrast to the verb o- in (13) the verbttena- in (8) can have its ablative NP promoted to a DO in order to facilitatethe application of the gap strategy. In order to remedy the ungrammatical

10 The conjunctive marker -ase is sometimes abbreviated to -a, particularly in informal Korean(Lukoff (1982: 159)).

205

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

JAE JUNG SONG

relative clause in (13.a), the verb ttena- is introduced and further conjoinedwith the original verb o- using the conjunctive marker -ase. This way theablative NP in question becomes associated with the newly introduced verb,under the licence of which it can be further promoted to a DO as well. To putit differently, the verb that can have ablative NPs promoted to DOs in favorof the gapping relativization is recruited to form a conjoined constructionwith the original verb, which does not have this ability. Note that thecombination of these two verbs does not bring about any 'disruptive'semantic change. The original verb o- meaning 'to come' can be combinedwith the added verb ttena- meaning 'to leave', because the semantics of theformer verb is compatible with the semantics of the latter verb, or vice versa:one has to leave one place in order to come to a different place, (or if oneleaves one place, one necessarily comes to a different place).

Being promoted to a DO via the collaboration of the conjunction andpromotion devices, the original ablative NP can now undergo the gapstrategy. So the ungrammatical relative clause in (13.a) can be changed intoa grammatical one:

(22) [kiho-ka ttena-ase o-n] hocuKeeho-SU leave-CONJ come-REL] Australia'(lit.) Australia, which Keeho left and came (here)'

The relative clause in (22) explicitly indicates that Australia is the place thatthe actor left.

Further, note that (22) has an alternative reading: it also means 'Australia,for which Keeho left (and to which he came)'. Why is (22) then ambiguous?Since the conjunction construction contains the two verbs (i.e. theoriginalverb o- under the licence of which the allative NP can be promoted to a DO;and the newly introduced verb ttena- under the licence of which the ablativeNP can be promoted to a DO), the head NP can be construed or associatedwith either the allative or ablative role in the relative clause. In other words,the relative clause in (22) is related to the two non-relative clauses:

(23) kiho-ka hocu-lul ttena-ase o-ass-taKeeho-SU Australia-DO leave-CONJ come-PST-IND'Keeho left Australia and came (here).'

(24) kiho-ka ttena-ase hocu-lul o-ass-taKeeho-SU leave-CONJ Australia-DO come-PST-IND'Keeho left (there) and came to Australia.'

In (23) and (24) the position of the promoted NP is fixed and placedimmediately before the verb that licenses the promotion; there is no room forambiguity. However, in the process of the gap relativization, this positionalfeature along with the NP and its postposition is obliterated (' '

206

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

KOREAN RELATIVE CLAUSE CONSTRUCTIONS

represents a deletion site; (25) corresponds to (23), whereas (26)corresponds to (24)):

(25) [kiho-ka ttena-ase o-n] hocuKeeho-SU leave-CONJ come-REL] Australia'Keeho left Australia and came (here).'

(26) [kiho-ka ttena-ase o-n] hocuKeeho-SU leave-CONJ come-REL Australia'Keeho left (there) and came to Australia.'

(25) and (26) are exactly the same on the surface, since the deletion sites arenot perceptible. This explains the potential ambiguity of the relative clause in(22). H

Now consider the ungrammatical relative clause in (14.a) again:

(14) ALLATTVEa. *[kiho-ka ttena-n] hocu

Keeho-SU leave-REL Australia'Australia, for which Keeho left'

The ungrammaticality of (14.a) is due to the failure of the allative NP to bepromoted to a DO with respect to the verb ttena-. In order to render (14.a)grammatical, the conjunction device is again called for: this time the verb o-,which allows allative NPs to be promoted to DOs, can be introduced to beconjoined with the original verb in (14). And again (22) (or (26)) above willbe exactly the output of the conspiracy between the promotion and Aconjunction devices, and the gap strategy.

The following relative clause also contains a verb that can promote onlyallative NPs to DOs; with ka- 'to go', ablative NPs cannot directly berelativized via the gap strategy (in other words, (27) is grammatical only if itmeans 'that place to which Keeho went'):

(27) * [kiho-ka ka-n] ku-kosKeeho-SU go-REL that-place'that place from which Keeho went'

11 One may wonder why this kind of ambiguity is tolerated in relative clauses when the promotiondevice itself discriminately applies to either allative or ablative in order to avoid any semantic roleconfusion. To put it in a different way, what is it that allows the A conjunction device tointroduce the additional verb into the relative clause only to cause ambiguity? It seems to me that,although it is imperceptible on the surface, the exact deletion site within the relative clause iseasily retrievable. In contrast to the non-relative clause, the relative clause codes a proposition (orinformation) that the speaker assumes is known or accessible to the hearer (Givón (1990:646)):the speaker uses the relative clause to help the hearer to identify the referent. Since the propositioncoded by the relative clause is mutually accessible, relative clauses such as in (22) are not really'ambiguous enough' to cause the semantic role confusion.

207

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

JAE JUNG SONG

The conjunction device is called into service to render-(27) grammatical; theverb ttena- can now promote the ablative NP in question to a DO so that itcan undergo the gap strategy:

(28) [kiho-ka ttena-ase ka-n] Icu-kosKeeho-SU left-CONJ go-REL that-place'that place from which Keeho went'

Note that, as with (22), (28) has an additional reading: the place to whichKeeho went. For the same reason explicated above, the relative clause isambiguous: the head NP can be associated with either of the two verbscontained in (28).

Now the K conjunction device, which I am about to describe, is slightlydifferent from the A conjunction device in that the former does not involvethe promotion device. However, they are functionally equivalent in that byway of these devices the oblique NP in question 'turns up' as a DO so as tobe directly relativized.

Consider (15) again:

(15) INSTRUMENTALa. *[kiho-ka ka-n] ku cha

Keeho-SU go-REL the car'the car in which Keeho went'

b. kiho-ka ku cha-lo ka-ass-taKeeho-SU the car-INST go-PST-IND'Keeho went in the car.' • ,-•

This particular example was used to illustrate not all instrumental NPs can berelativized. (15.a) is the result of the gap strategy applied to the grammaticalsentence in (15.b) with ku cha 'the car' as the head NP.

Is there any other way in which Korean can express what theungrammatical relative clause in (15.a) intends to express, i.e. the head NPwhich has an instrumental function within the relative clause? To put it in ageneral way, is there any other way that the functional communicativerequirement of which Croft (1990) speaks can be met? It seems that there is.

Korean has two ways of expressing certain oblique relations, i.e.instrumental, and comitative: (a) use of explicit postpositions; or (b) use ofseparate verbs that may indirectly denote these oblique relations. Forinstance, the postposition, -lo, marks the instrumental role, as in:

(29) kiho-ka khal-lo koki-lul ccal-ess-taKeeho-SU knife-INST meat-DO cut-PST-IND'Keeho cut the meat with a knife.'

208

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

KOREAN RELATIVE CLAUSE CONSTRUCTIONS

Alternatively, the same case relation can be periphrastically expressed viathe K conjunction device:12

(30) kiho-ka khal-ul kaci-ko koki-lul ccal-ess-taKeeho-SU knife-DO take-CONJ meat-DO cut-PST-IND'(lit.) Keeho took a knife and cut the meat.'

The above sentence is an example of a conjoined sentence, as indicated bythe conjunctive marker -ko linking the two clauses. In addition to its usualfunction of coding temporally ordered (not necessarily related) actions, theconjunctive marker -ko is also used to express the adverbial function of thefirst conjunct in relation to the second (Lukoff:1982:104-105,364).i3 Infact, (30) is better translated as 'Keeho cut the meat, using (or holding) theknife' (given the pragmatic knowledge that people use knives to cut meat).Note that in (30) the NP khal 'a knife' is no longer an instrumental NP, but aDO NP of the first or additional verb, the semantics of which successfullyretains the meaning of the instrumental role of the NP in question.14

And precisely this alternative way of expressing the instrumental relationis exploited in turning the ungrammatical relative clause in (15.a) into agrammatical one. Consider the grammatical counterpart of (15.a):

(31) [kiho-ka tha-ko ka-n] ku chaKeeho-SU ride-CONJ go-REL the car'(lit.) the car that Keeho got on and went'

12 I am not saying that other languages cannot do this; in fact, I think any language can expressoblique relations in either way. The point that I am about to make here is that the conjunctiondevice in (b) is systematically used: (1) to express adverbial functions (see below); and (2) toparticipate in the relativization conspiracy in Korean.

13 For instance, the following sentence can be interpreted as either a sequence of two temporallyordered separate actions or connected actions.(i) kiho-ka campa-lul ip-ko cip-ul naka-ass-ta

Keeho-SU jumper-DO wear-CONJ house-DO leave-PST-IND(a) 'Keeho put on his jumper and left the house.'(b) 'Keeho left the house, wearing his jumper.'The first translation in (i) represents two temporally ordered separate actions, whereas the secondtranslation in (i) represents two connected actions. And the second reading is what Lukoff has inmind when he (1982:364) speaks of the -ko conjunctive clause functioning as adverbialexpressions: the first conjunct effectively describes the manner in which the actor left the house.When inserted in (i), the manner adverb will have only a local scope in the first reading. On theother hand, the same adverb will have a global scope in the second reading, i.e. when the conjoinedconstruction 'functions as expressing an adverbial expression'. Cf. :(ii) kiho-ka coyonghi campa-lul ip-ko cip-ul naka-ass-ta

Keeho-SU quietly jumper-DO wear-CONJ house-DO leave-PST-IND(a) 'Keeho quietly put on his jumper and left the house.'(b) 'Keeho quietly left the house, wearing his jumper.'The first reading indicates that Keeho was quiet only when he was putting on his jumper, whereasthe second reading indicates that Keeho was quiet when he left the house with his jumper on. Andit is the adverbial function of the K conjunction device that is put to use in the relativizationconspiracy.

14 I would like to add that the alternative way of marking the instrumental role is very reminiscent ofserial verb languages (e.g. Chinese, Akan, Ijo, Yoruba etc) in which case relations are expressed byseparate verbs instead of case markers such as adpositions (see Sebba (1987), Li & Thompson(1973), Lord (1973), Ansre (1966), inter alia).

209

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

JAE JUNG SONG

The additional verb tha- 'to ride' is used in conjunction with the verb ka-\ theverb that has taken over the role of marking the instrumental NP is conjoinedwith the original verb using the conjunctive marker -ko. More importantly,now that it has become a DO NP of the newly introduced verb, the NP kucha 'the car' can undergo the primary gap RC forming strategy without anydifficulty. Recall that the gap strategy freely relativizes DOs in Korean.Evidence that the NP ku cha has become a DO comes from the followingnon-relative clause (as signalled by the DO postposition that it now bears):15

15 Some may point out that in (i) the verb tha- appears with a locative NP, and that in (ii) the NPku cha 'the car' in fact is a locative NP that has been promoted to a DO:(i) kiho-ka ku cha-ey tha-ass-ta

Keeho-SU the car-LOC ride-PST-IND'Keeho rode in the car.'

(ii) kiho-ka ku cha-lul tha-ass-taKeeho-SU the car-DO ride-PST-IND'Keeho rode in the car.'

In other words, the promotion of the locative NP to the DO is claimed to account for therelationship between (i) and (ii). However, there is evidence that the verb tha- is both intransitiveand transitive: the verb in (i) is an intransitive requiring no DO, whereas the 'same' verb in (ii) is atransitive verb subcategorizing a DO NP. First, those verbs that can promote oblique NPsconsistently promote proforms for oblique NPs. For instance, the verb o-, which has been shownto promote oblique NPs to DOs, also promotes the proform oblique NP (note that thepostposition ey marks both locative and allative NPs):(iii) kiho-ka yeki-ey o-ass-ta

Keeho-SU this=place-ALL come-PST-IND'Keeho came to this place (=here).'

(iii1) kiho-ka yeki-lul o-ass-taKeeho-SU this=place-DO come-PST-IND'Keeho came to this place (=here).'

The above examples support the view that the promoted proform NP still retains the originalmeaning of the oblique role: (iii) and (iii') are related to each other via the promotion device.On the other hand, the verb in question tha- in (i) and (ii), behaves syntactically quits differentlyfrom the promotion verb. So the verb tha- in (i) can take the proform locative NP:(iv) kiho-ka yeki-ey tha-ass-ta

Keeho-SU this=place-LOC ride-PST-IND'Keeho rode in this place (=here).'

In contrast, the same verb in (ii) cannot have any proform locative NP marked as a DO:(v) *kiho-ka yeki-lul tha-ass-ta

Keeho-SU lhis=place-DO ride-PST-IND'*Kecho rode this place (=here).'

This difference in syntactic behavior demonstrates that the NP ku cha in (ii) is a genuine DO NPof the transitive verb: it is not a locative NP that has been promoted to a DO.Further evidence is furnished by the distribution of different WH-words with respect to thepromotion verb like o- and the transitive verb like tha-. One can never use the WH-word mues'what', when it occurs in the position that the promoted allative NP occupies:(vi) *kiho-ka mues-ul o-ass-ni

Keeho-SU what-DO come-PST-Q'What (place) did Keeho come to?'

Instead, one must always use the locative WH-word, eti:(vii) kiho-ka eti-lul o-ass-ni

Keeho-SU where-DO come-PST-Q'Where did Keeho come?'

But with respect to the transitive verb in question, the WH-word mues is fully acceptable:(viii) kiho-ka mues-ul tha-ass-ni

Keeho-SU what-DO ride-PST-Q'What did Keeho ride?'

210

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

KOREAN RELATIVE CLAUSE CONSTRUCTIONS

(32) kiho-ka ku cha-lul tha-ko ka-ass-taKeeho-SU the car-DO ride-CONJ go-PST-IND'Keeho got on the car and went.'

Another example of the K conjunction device exploited in the relativizationconspiracy is:

(33) [kiho-ka tul-ko yenghi-eykey kil-ulKeeho-SU hold-CONJ Yonghee-IO way-DO

annayha-n] chospulshow-REL candlelight

"The candlelight with which Keeho showed the way to Yonghee.'

Compare (33) with the ungrammatical relative clause produced when thisdevice is not used:

(34) *[kiho-ka yenghi-eykey kil-ul annayha-n] chospulKeeho-SU Yonghee-IO way-DO show-REL candlelight'the candlelight with which Keeho showed the way to Yonghee.'

Note that the relationship between the actor and the instrument is taken intoaccount in order to select an appropriate verb, e.g. ml- 'to take' vs tha- 'toride'. And this particular device seems to be restricted to the instrumental NP(and the comitative NP, as will be shown below).

It is clear that the A and K conjunction devices under discussion arefunctionally equivalent to the grammatical device in Toba Batak that promotesDOs to SUs (or the device in Korean that promotes allative or ablative NPs toDOs) so that they can be relativized via the primary strategy. When theoriginal verb cannot promote an oblique NP to a DO for the gap strategy, theA conjunction device introduces a verb which is semantically compatible withthe original verb, and which also can perform the necessary promotion.Similarly, the.K conjunction device introduces a verb which reflects themeaning of the postposition eliminated along with the relativized NP, andwhich at the same time subcategorizes the NP as a DO, a position that iseasily accessible to the gap RC forming strategy.

On the other hand, the locative WH-word is not possible with this verb:(ix) *kiho-ka eli-lul lha-ass-ni

Keeho-SU, where-DO ride-PST-Q'(lit.) Where did Keeho ride (or get on)?'

In sum, the foregoing syntactic evidence argues against the promotion analysis of (ii): the verb in(ii) is a transitive one in its own right; and the same verb in (i) is an intransitive verb. The verbtha- must be listed as both an intransitive and transitive verb in the lexicon, whereas the verb o-must be listed only as an intransitive verb.

211

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

JAE JUNG SONG

3.1.3 Adverb Device

The last device that I will examine can be demonstrated with reference to theungrammatical relative clause in (16):

(16) COMTTATTVEa. *[kiho-ka san-ey ka-n] sunay

Keeho-SU mountain-ALL go-REL Sunae'Sunae, with whom Keeho went to the mountain'

b. kiho-ka sunay-wa san-ey ka-ass-taKeeho-SU Sunae-COM mountain-ALL go-PST-IND'Keeho went to the mountain with Sunae.'

Here adverbs such as kathi or hamkkey, both meaning 'together', can beused to make up for the loss of the comitative postposition that is caused bythe gapping of the relativized NP:

(35) [kiho-ka kathilhamkkey san-ey ka-n] sunayKeeho-SU together mountain-ALL go-REL Sunae'Sunae, with whom Keeho went to the mountain'

Clearly, the meaning of the comitative case marker eliminated along with theNP in the gap relativization is restored via the adverbs in question: thepresence of either of the adverbs works with the gap RC forming strategy tothe effect that the otherwise unrelativizable comitative NP can be relativizedwithout any difficulty. And this particular device seems to be used only withrespect to comitative NPs that cannot be directly relativized using the gapstrategy.

Incidentally, the ungrammatical relative clause in (16) can also be turnedinto a grammatical one by way of the K conjunction device. Consider:

(36) kiho-ka sunay-lul teyli-ko san-eyKeeho-SU Sunae-DO accompany-CONJ mountain-ALL

ka-ass-tago-PST-IND

'Keeho accompanied Sunae and went to the mountain.'

Again, in (36) the original comitative NP has become a DO (of the newlyintroduced verb teyli-) after the operation of the K conjunction device; it cannow undergo the gap strategy, producing the grammatical relative clause:

(37) [kiho-ka teyli-ko san-ey ka-n] sunayKeeho-SU accompany-CONJ mountain-ALL go-REL Sunae'Sunae, whom Keeho accompanied and went to the mountain'

212

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

KOREAN RELATIVE CLAUSE CONSTRUCTIONS

3.1.4 Summary

Keenan and Comrie (1977) demonstrate, among other things, that devicesalready available in grammar are ingeniously exploited in order to enable RCforming strategies to relativize NPs that they otherwise cannot. Three suchdevices have been identified as being involved in the Korean relativizationconspiracy: promotion, conjunction (A and K types), and adverbs: recoursemust be had to these devices in order to successfully relativize otherwiseunrelativizable oblique NPs. These devices in Korean are functionallyequivalent to the promotional device used in the relativization conspiracy inToba Batak, although the adverb device may seem to be very 'roundabout' incomparison with the other devices.

Finally, in Korean, contrary to Keenan and Comrie (1977), the primarystrategy does not always apply to OBL (except for LOCs); where the gapstrategy fails, one or two of the conspiratorial devices must be called into therelativization conspiracy.

4. THE RC FORMING STRATEGIES AND PRAGMATICS

Why is it the case that some oblique NPs can be directly relativized withoutrecourse to one of the conspiratorial devices, while other oblique NPs canonly be relativized using one of the devices?

As a starting point, consider the relative clause in (2) again:

(2) [kiho-ka ku kay-lul ttayli-n] maktaykiKeeho-SU the dog-DO beat-REL stick'the stick with which Keeho beat the dog'

The postposition eliminated along with the relativized NP is the instrumentalmarker, -lo. And this postposition is the only structural element thatindicates the instrument relation between the action of beating the dog and thestick. What is it then that unambiguously helps retrieve the instrumentrelation in the construal of (2)?

Comrie (1981:145) briefly discusses how the instrumental interpretationassociated with the gapped NP in (2), i.e. maktayki 'the stick', is recoveredor retrieved. Comrie correctly points out that it is common sense (orpragmatics) against which the 'missing' relation between the actor's actionand the NP in question is construed: in the real world the most likely relationbetween the act of someone's hitting a dog and a stick is that of instrument.16

That this is indeed the way how common sense plays such an important rolein the recovery of the missing relation is illustrated very well by the

16 This is possible to a marginal extent in English, e.g. I've got no place to live (cf. I've got no placeto live in) or Give me wings to fly (cf. Give me wings to fly with). Matsumoto (1989:233)provides more attested English examples of this kind, e.g. Here is a snack that you don't have towash your hands (cf. Here is a snack that you don't have to wash your hands in order to eat).

213

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

JAE JUNG SONG

following relative clause, quite similar to (2), except for that the relativizedNP is paychu 'the cabbage', instead of maktayki 'the stick':

(38) V.[Mho-ka ku kay-lul ttayli-n] paychuKeeho-SU the dog-DO beat-REL cabbage'the cabbage with which Keeho beat the dog'

It is extremely difficult to interpret (38) (as indicated by the preceding doublequestion marks), since in the real world people normally do not beat a dogwith a head of cabbage.17 In fact, this is why the K conjunction device iscalled for in order to render (38) fully grammatical or more natural:

(39) [kiho-ka tul-ko ku kay-lul ttayli-n] paychuKeeho-SU hold-CONJ the dog-DO beat-REL cabbage'Keeho held the cabbage and hit the dog (with it).'

Further compare (10) with (16):

(10) COMITATTVEa. [kiho-ka kyelhonha-n] ku yein

Keeho-SU marry-REL the lady'the lady whom Keeho married'

b. kiho-ka ku yein-kwa kyelhonha-ess-taKeeho-SU the lady-COM marry-PST-IND'Keeho married the lady.'

(16) COMTTATTVEa. *[kiho-ka san-ey ka-n] sunay

Keeho-SU mountain-ALL go-REL Sunae'Sunae, with whom Keeho went to the mountain'

b. kiho-ka sunay-wa san-ey ka-ass-taKeeho-SU Sunae-COM mountain-ALL go-PST-IND'Keeho went to the mountain with Sunae.'

In both relative clauses in (lO.a) and (16.a) the relativized NPs are comitativeNPs. But why is there a difference in grammaticality? In the case of (lO.a)

17 One of the anonymous AJL referees points out that it may be the verb within the relative clauserather than pragmatics that seems more responsible for the oddity of (38): by virtue of theirsemantic structure different verbs treat OBLs differently. For instance, an INST may be moreimportant to the verbs like ttayli- 'to hit', mandul- 'to make' etc than the verbs like annayha- 'toshow', manna- 'to meet' etc. However, if it were indeed the semantic structure of a verb thatdetermines whether a given OBL NP can be directly relativized via the gap strategy, one wouldexpect a given verb to behave consistently with respect to relativization. As the rest of my paperclearly shows, ungrammatical or odd sentences (e.g. (42), (43)) will sound natural given the rightcontext. Further, I do not believe that anyone has ever succeeded in making a clearcut distinctionbetween semantics and pragmatics. Nor do I believe that it is possible to distinguish semanticsfrom pragmatics. The semantic structure of a verb does not arise ex nihilo. The verb meaning is,like other kinds of meaning, largely derivative of the way how human beings construe the activitydenoted by the verb.

214

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

KOREAN RELATIVE CLAUSE CONSTRUCTIONS

pragmatic knowledge of marriage helps retrieve the comitative relationeliminated by the gap strategy. One cannot marry oneself. It takes anotherhuman partner (of the opposite sex) to constitute a marriage. Marriage, then,necessarily involves a comitative relation. This piece of pragmaticinformation is what makes (lO.a) acceptable.18 On the other hand, in thecase of (16.a) no such obvious pragmatic knowledge is available. Going to amountain does not necessarily involve à partner: one can go to a mountainalone. Therefore, there is no comparable pragmatic knowledge that helpsrecover the comitative relation erased by the gap strategy.

Generally, it is this kind of pragmatic knowledge that ultimatelydetermines whether a given oblique NP can be directly relativized via the gapstrategy.19 When the. relationship between the head NP and the action orprocess denoted by the predicate or verb can pragmatically be established orretrieved, the gap strategy can be applied without any complication. Whensuch pragmatic inferencing is difficult, or even impossible, one of thegrammatical devices is called into the relativization conspiracy. This way therelation expressed by the postposition that was eliminated by the gap strategyis then effectively revived, albeit indirectly.

Matsumoto (1990) carries out an in-depth study of the role of pragmaticsin the Japanese relative clause, which is structurally very similar to theKorean relative clause. Matsumoto relies on the frame semantics of Fillmore(1982) and 'world-view' (cf. Ross (1975)) in order to account for some'anomalous' facts about Japanese relative clauses.20 For instance,Matsumoto (1990:113-114) notes that no apparent coreferential NPs areinvolved in some Japanese grammatical relative clauses:

(40) JAPANESE[genki no derü] terebi ga mitaienergy NOM rise TV NOM want-to-watch'( ) want(s) to watch a TV program (by watching which) (one's)energy will rise.'21

18 In Korean the verb kyelhonha- is an intransitive verb that does not require any complement. Hencethe comitative relation in question is not either syntactic or semantic information.

19 Of course, this generalization does not apply to those syntactic positions on AH that the strategiesfreely apply to, i.e. SU, DO, IO, LOC, and GEN.

20 Roughly, Fillmorean frame semantics advocates the idea that a linguistic element is interpreted inthe larger context which it evokes (Matsumoto 1990:116-117). In other words, a description of alinguistic element includes all possible participants associated with it; e.g. a commercial eventwould include the buyer, the seller, goods, money, the place, purpose, beneficiary, etc. The world-view that Matsumoto speaks of here is derived from Ross (1975), who refers to the culturally-based knowledge of the world that allows one to predict interpretations and relationships regardingnew information, events, and experiences (Matsumoto 1989:228). Although she tries to draw adistinction between frame and world-view, I have found it very difficult to appreciate Matsumoto'sdistinction. I think that it is safe for purposes of this paper to include frame and world-view underthe heading of pragmatics or extra-linguistic knowledge (as indeed suggested by the title ofMatsumoto (1990)).

21 The parentheses in the translation, as in (40), represent the argument/NP that is 'dropped' due tothe presence of contextual information. As will be shown below (see also note 22), Korean sharesthis characteristic.

215

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

."• JAE JUNG SONG

Note that the verb in the relative clause is intransitive, and that there is nogapped argument for the relativized position to be referentially bound to(Matsumoto 1990:117-118). Matsumoto goes on to argue that the world-view or background knowledge of the hearer (i.e. pragmatics) is such thatthe watching of certain television programs is a possible source of vigor orenergy. And this extra-linguistic information will enable one to construe therelationship between the head NP and the relative clause.

I will not here discuss Japanese relative clauses and the role of pragmaticstherein. Rather, I will question what Matsumoto has to say about Koreanrelative clauses on the basis of Na (1986); I will argue that the role ofpragmatics in Korean relative clauses is underestimated in the work of bothMatsumoto and Na.

While trying to determine possible constraints influencing acceptability ofrelative clauses, Matsumoto (1990:120) entertains the notion of anaccessibility hierarchy, such as, proposed by Na (1986) in her analysis ofKorean relative clauses. Na (1986:140) claims:22

(41) Arguments must be filled by the head NP of a RC in thefollowing order: {subcategorized [i.e. syntactic -JJS]elements > semantically entailed elements > pragmaticallyentailed elements}. An argument position lower in thehierarchy may not be filled when there is an unfilledargument position that is higher in [the] hierarchy.

Na argues that this explains why the following Korean relative clause isungrammatical:

(42) *[Ann-i kkakke cu-(u)n] salamAnn-SU peel give-REL person'the person (for whom) Ann peeled (something)'

According to Na (1986:139), the verb sequence kkakk-e cu- subcategorizesthe direct object. However, in (42) this position is not filled; instead, herargument goes, the beneficiary NP, salam, which is only semanticallyentailed by the composite verb, has been filled by the head NP.23 In Na'stheory, this is a clear violation of the accessibility hierarchy in (41). Hencethe alleged ungrammaticality of (42).

22 As in Japanese (Matsumoto 1990:113), in Korean core arguments can remain unexpressed or canbe 'dropped' even in a non-relative sentence, if they are recoverable from the context. Forexample, ( ) ( ) ttayliessta '( ) hit ( )' is a fully grammatical sentence, even though it containsno arguments. Therefore, the hierarchy in (41) is intended to predict the order in which 'missing'arguments including relativized/gapped NPs can be referenlially bound to head NPs, when clausescontaining such missing arguments as in the above example go through the relativization process.

23 The construction used in (42) is that of the conjunction construction discussed in section 3.1.2,i.e. the A conjunction device. And the beneficiary NP in (42) in fact is not, pace Na, asemantically entailed element; it is the subcategorized argument of the second verb cu- in theconjunction construction. Similarly, in (43) it is, in fact, a case of a pragmatically entailedelement gaining priority over a subcategorized argument.

216

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

KOREAN RELATIVE CLAUSE CONSTRUCTIONS

Na (1986:140) further illustrates the order between semantically andpragmatically entailed elements in Korean relative clauses by way of the nearunacceptability of:

(43) T?[Alice-ka sakwa-lul kkakk-e cu-(u)n] khalAlice-SU apple-DO peel-CONJ give-REL knife'the knife (with) which Alice peeled an apple (for X)'

The reason why (43) is almost unacceptable is claimed to be that thesemantically entailed element (i.e. for X) remains empty, while the head is apragmatically entailed element (i.e. the instrument relation between the act ofpeeling apples and the knife is pragmatically inferred; see the discussion on(2) and (38) above).

Matsumoto (1990:121) compares her own analysis of Japanese relativeclauses with Na's analysis of Korean relative clauses, and abandons thenotion of the accessibility hierarchy as in (41) for Japanese, since '[although]the proposed hierarchy is systematic, and appears plausible [with respect toKorean],... it may not be extendable without modifications to Japanese ...[Therefore, it] seems to be too strong for Japanese.' However, I will nowdemonstrate that Na's hierarchy is also too rigid or idealized to account forKorean relative clauses.

Consider (42) again. It is fully acceptable given the right context.Suppose that people are engaged in some novel game of peeling for theirfriends as many apples as possible in a given span of time. Then it is sharedpragmatic knowledge that apples are being peeled in the game. Given thiscontext, one can say with full naturalness:

(44) [Ann-i kkakk-e cu-(u)nj salam-un eâ ka-ass-ni. Ann-SU peel-CONJ give-REL person-TOP where go-PST-Q

'Where has the man for whom Ann peeled (apples) gone?'

In (44) the 'semantically entailed' element is construed with the head NP,leaving the subcategorized argument unfilled. This is a clear violation ofNa's hierarchy.

As for (43), it is also acceptable, given the right context: suppose that thespeaker is asking where the knife with which Alice used to peel apples forthe hearer last time:

(45) [cinanpen Alice-ka sakwa-lul kkakk-e cu-(u)n]last-time Alice-SU apple-DO peel-CONJ give-REL

khal-ul etita tu-ess-niknife-DO where place-PST-Q

'Where did you put the knife with which Alice peeled apples (foryou) last time?'

217

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

JAE JUNG SONG

In (45) the speaker is appealing to the shared pragmatic knowledge that it isfor the hearer (or the semantically entailed beneficiary argument in Na'sterms) that Alice peeled apples last time. (45) is then a case of apragmatically entailed element (i.e. the relationship between the act of peelingapples and the knife) gaining priority over a semantically entailed one (i.e.the beneficiary argument). Clearly, this is also in violation of Na's hierarchy.

Matsumoto (1990:121-124) adduces additional examples which plainlymilitate against Na's hierarchy in an attempt to demonstrate that, despitesome structural resemblances between them, the role of pragmatics in theJapanese relative clause is really far more significant than in the Koreanrelative clause. In what follows, I will provide parallel examples in Koreanto prove that Korean is not different from Japanese after all, insofar as therole of pragmatics in relative clauses is concerned.

Consider the following (the sentences are Korean equivalents toM(atsumoto)'s Japanese examples, the actual numbers of which appear in theparentheses):

(46) = (M's (16))[sa-l] ton-i iss-nibuy-REL money-SU exist-Q'(lit.) Do (you) have money (with which) (you will) buy (it/one)?''Do you have money to buy (it/one)?'

(47) = (M's (17))[mek-un] kulus-un kkaykkusi ttakk-a-laate-REL bowl-TOP neatly wash-PF-IMP'Wash the bowls (from which) (you) have eaten'

(48) = (M's (18))[ecey mek-un] siktang-un meyu pumpi-ess-tayesterday ate-REL restaurant-TOP very crowded-PST-IND'The restaurant (at which) ( ) ate yesterday was very crowded.'

The above examples clearly demonstrate that Na's hierarchy is as strong forKorean as for Japanese: in all the above cases the pragmatically entailedelements are chosen over the subcategorized elements, and referentiallybound to the head NPs. For instance, in (47) it is the pragmatic knowledgeabout the activity of eating, and the role of a bowl in the activity that links thehead NP, kulus 'the bowl', and the predicate, mek-'ate', to the effect that thepostposition eliminated along with the NP is construed or retrieved as that ofinstrument. In this context, note that the other missing argument is noneother than the subcategorized argument of the predicate in question, i.e. thesubject NP.

Further, as in Japanese (e.g. (40)), in Korean some relative clauses do noteven involve coreferential NPs. I will provide one such example (taken fromTagashira (1972); for more, see Tagashira (1972:224-225)):

218

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

KOREAN RELATIVE CLAUSE CONSTRUCTIONS

(49) [kwika-ka nuceci-nun] kyowoysaynghwalgetting=home becoming=late-REL suburban life'?the suburban living of which the getting home becomes late(Tagashira's translation)'

The pragmatic knowledge involved in the generation and construal of (49)may be that it increasingly takes more time to travel from a downtownworkplace to a suburban house than to a downtown house and/or that moreand more people move out to live in the suburbs, contributing to trafficcongestion in between. And it is this extra-linguistic knowledge that makesup for the lack of any coreferential NP in (49).

I conclude that what clearly determines the most likely interpretation of agiven relative clause in Korean is not such a strictly ordered set of elements(i.e. subcategorized, semantically, or pragmatically entailed; the accessibilityhierarchy proposed by Na (1986)), but such pragmatic factors as identifiedfor Japanese relative clauses by Matsumoto (1990) (i.e. context, backgroundknowledge, or pragmatic knowledge; Matsumoto (1990:124) also reaches thesame conclusion for Japanese). Contrary to Matsumoto and Na, Korean isnot so different from Japanese, insofar as the role of pragmatics in relativeclauses is concerned.

5. CONCLUSION

In this paper, I have argued that the Korean relative clause construction is notas simple or straightforward as Keenan and Comrie (1977) characterize it intheir cross-linguistic research. In particular, I have demonstrated: (a) that thegapstrategy does not always apply to oblique NPs by itself; and (b) that thegap strategy also applies to some genitive NPs.

I have also discussed in detail the three devices that are pressed into theconspiracy with the primary RC forming strategy in Korean. One (or two inthe case of the A conjunction device) of these devices is called for so thatotherwise unrelativizable oblique NP can be ultimately relativized via the gapRC forming strategy.

Finally, I have argued that the accessibility hierarchy espoused in Na(1986) is as strong for Korean relative clauses as for Japanese relativeclauses. Pragmatics - context, background knowledge, and world-view -plays as important a role in the generation and construal of relative clauses inKorean as Matsumoto (1990) has demonstrated for Japanese relative clauses.

REFERENCES

Ansre, G. 1966. The verbid - a caveat to 'serial verbs'. Journal of West AfricanLanguages 3. 29-32.

Comrie, B. 1975. Causatives and universal grammar. Transactions of the PhilologicalSociety 1974. 1-32.

219

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012

JAE JUNG SONG

Comrie, B. 1976. The syntax of causative constructions: cross-language similarities anddivergences. In Shibatani, M. (ed), Syntax and semantics vol. 6: the grammar ofcausative constructions. New York: Academic Press. 261-312.

Comrie, B. 1981. Language universals and linguistic typology. Oxford: Blackwell.Croft, W. 1990. Typology and universals. Cambridge: Cambridge University Press.Fillmore, C.J. 1982. Frame semantics. In: Linguistics Society of Korea, (ed),

Linguistics in the Morning Calm. Seoul: Hanshin. 111-138.Foley, W.A. & Van Valin, R.D. 1984. Functional syntax and universal grammar.

Cambridge: Cambridge University Press.Givón, T. 1990. Syntax: a functional-typological introduction, volume II. Amsterdam:

John Benjamins.Greenberg, J.H. 1966. Some universals of grammar with particular reference to the order

of meaningful elements. In Greenberg, J.H. (ed), Universals of grammar.Cambridge, MA: MIT Press. 73-113.

Keenan, E. & Comrie, B. 1977. Noun phrase accessibility and universal grammar. LI 8.63-99.

Keenan, E. & Comrie, B. 1979. Data on the noun phrase accessibility hierarchy. Lg 55.333-51.

Li, C.N. & Thompson, S.A. 1973. Co-verbs in Mandarin Chinese: verbs orprepositions? Journal of Chinese Linguistics 2. 257-278.

Lord, C. 1973. Serial verbs in transition. Studies in African Linguistics 4. 269-296.Lukoff, F. 1982. An introductory course in Korean. Seoul: Yonsei University Press.Mallinson, G. & Blake, B. 1981. Language typology. Amsterdam: North-Holland.Matsumoto, Y. 1989. Japanese-style noun modification ... in English. BLS 15.

226-37.Matsumoto, Y. 1990. The role of pragmatics in Japanese relative clause constructions.

Lingua 82. 111-29.Na, Y. 1986. Syntactic and semantic interaction in Korean: theme, topic, and relative

clauses. Unpublished Ph.D. dissertation, University of Chicago.Ross, R.N. 1975. Ellipsis and the structure of expectation. San Jose State Occasional

Papers in Linguistics 1. 183-191.Sebba, M. 1987. The syntax of serial verbs. Amsterdam: John Benjamins.Song, J.J. 1991. Causatives and universal grammar: an alternative interpretation.

Transactions of the Philological Society 89. 65-94.Tagashira, Y. 1972. Relative clauses in Korean. In: Peranteau, P.M., Levi, J.N. &

Phares, G.C. (eds), The Chicago which hunt: Papers from the relative clause festival.Chicago: Chicago Linguistic Society. 215-229.

LinguisticsEnglish DepartmentUniversity of OtagoP.O.Box 56DunedinNew Zealand

220

Dow

nloa

ded

by [

Uni

vers

ity o

f A

rizo

na]

at 0

2:21

09

Aug

ust 2

012